I have long supported old-style platform immunity protecting providers from liability for the content of a search result, forum post, or the like.

Both individually tailored content recommendation algorithms and LLM output are different, and companies should be liable for the output of both when they're used in the way social media platforms and search engines are currently using them.